04 · What Is Storytelling? | Generative AI for Game Development
This chapter companion breaks game storytelling into plot, characters, setting, conflict and resolution, then shows how each gives generative AI the structure it needs. It closes with a short history of storytelling tools.
This video is a companion to the fourth chapter of Generative AI for Game Development by Shesh Narayan Gupta. It starts from the observation that people picture sharp dialogue when they think of game storytelling, then argues that dialogue is only the paint. The real work is the skeleton underneath, and that skeleton is what a developer needs to hand to an AI.
Why structure comes first
The video's central claim is that generative AI does not invent story logic. It studies thousands of stories and builds on the patterns it finds, so you give it the pattern first. The chapter's thesis is stated as: the stronger the narrative foundation, the better generative AI can support the story. The video lists four things structure unlocks: better prompts because you know what to ask for, more coherent output because the guardrails are narrative, better branching because you know the joints, and generated content that fits the story's world.
Plot
Plot is the sequence of meaningful events. The classical structure is exposition, rising action, climax and resolution. The video adds eight plot archetypes that keep recurring: the hero's journey, rags to riches, overcoming the monster, tragedy, voyage and return, rebirth, mystery, and transformation. For AI, plot structures are what generative models use to create quests and branching points that feel meaningful. Feeding the model the arc keeps its output on the rails.
Characters
Characters are the emotional anchor, and the video notes that players forgive a thin plot for a character they love. It lists seven roles, each with a job: protagonist, antagonist, companion, NPC, mentor, trickster, and the player-created avatar. The payoff for AI is that character attributes and motivations guide generated dialogue. Give the model a consistent personality and even its improvised lines stay in character.
Setting
Setting is treated as a character rather than a backdrop, since architecture, culture, weather and tavern rumors all tell a story. The video describes eight settings: fantasy, sci-fi, post-apocalyptic, historical, urban, horror, mythical and hybrid. Given a setting description, a model can generate lore entries, environmental narratives and procedural details that match the world's tone.
Conflict and resolution
Conflict is called the engine: without it there is no challenge, no growth and no reason to act. The eight types are character versus self, versus character, versus society, versus nature, versus technology, versus the supernatural, versus time, and versus the unknown. For AI, conflict patterns work as a quest generator. Asking for a versus-society conflict, for example, produces dilemmas where both answers cost something.
Resolution is the payoff but not always an answer. The seven types are closed, open-ended, branching, circular, tragic, player-defined and false. Knowing the type lets developers shape AI-generated endings that stay thematically consistent, because the model can improvise a last chapter once told what kind of ending it is.
Storytelling follows its tools
The video closes with a brief history: firelight gave us myths, writing gave us epics such as the Iliad and the Ramayana, theater added dialogue and an audience, film and television added image and edit, and games added agency. The AI chapter includes AI Dungeon, Project December, and mods that wire language models into Skyrim NPCs who remember what the player did. The video's conclusion is that AI does not replace the storyteller but expands what is possible, and that the principles hold whether a story is written by humans, machines or both.
Key takeaways
- Generative AI builds on narrative patterns; it does not invent story logic.
- Plot has a classical four-part arc and eight recurring archetypes.
- Seven character roles and consistent motivations keep generated dialogue in character.
- Setting acts as a prompt that keeps generated lore aligned with the world.
- Eight conflict types serve as a quest generator.
- Seven resolution types let developers steer generated endings.
- The tools change, but the job of moving, challenging and connecting players stays the same.
Try it yourself
Take the video's exercise: you are designing a game about a lighthouse keeper who finds a door that shouldn't exist. Pick the two of the eight plot archetypes that fit best and explain why, sketch a three-act structure, and name one quest an AI could generate that still fits your theme. The chapter belongs to the book Generative AI for Game Development, linked in the video description.
Full transcript(auto-generated, with timestamps)
[0:00]Generative AI for game development. What is storytelling? The building blocks every game story is made of and how to hand them to an AI. Hello, this is Liam in for Bear. When people hear game storytelling, they picture sharp dialogue. Dialogue is the paint. The real work is the skeleton underneath and that's what we're building today. Six terms then we're off. Plot, the sequence of events that moves a story. Character arc, how a character changes through conflict. Setting, the world where the story happens. Conflict, the tension that drives action. Resolution, how the story lands. An emergent narrative, stories that arise from play, not from a script. Generative AI doesn't
[0:37]Invent story logic. It studies thousands of stories and builds on the patterns it finds. So, if you want it to help you tell one, you give it the pattern first. That pattern is narrative structure, the skeleton every story hangs on. Four things structure unlocks. One, prompt the model better because you know what to ask for. Two, reducing coherent output because the guardrails are narrative. Three, design better branching because you know the joints. Four, generate content that actually fits your story's world. The chapter's thesis in one line, the stronger the narrative foundation, the better generative AI can support the story. Structure in, coherence out. Start with
[1:14]Plot, the sequence of meaningful events. The classical structure still rules. Exposition sets the world, rising action builds the stakes, the climax breaks them open, and the resolution lets everyone land. Then come the archetypes, eight plot shapes that keep showing up. The hero's journey, leave, transform, return. Rags to riches, rise from nothing. Overcoming the monster, face the terror and win. Tragedy, the flaw wins instead. Four more, voyage and return, there and back changed. Rebirth, a dark night then a new self. Mystery, the truth was buried. Transformation, the world itself changes shape. Why this matters for AI? Plot structures are what generative models use to create quests
[1:53]And branching points that feel meaningful. Feed the model the arc and the it generates stay on the rails. Characters are the emotional anchor. Players forgive a thin plot for a character they love and they'll abandon a brilliant one for a cast they don't. Motives, growth, flaws, that's the contract. Seven roles each with a job. The protagonist, the transformation engine, silent, defined or yours to build. The antagonist, the opposition, villain, mirror or the system itself. The companion, the emotional spine beside you. The NPC, the world alive around you. Three more, the mentor, wisdom before the fall. The trickster, chaos that teaches. And the player-created avatar, your face, your choices, your story.
[2:34]The AI payoff, character attributes and motivations are what guide generated dialogue. Give the model a consistent personality and every line it writes stays in character, even the improvised ones. Setting isn't the backdrop, it's a character. Architecture, culture, weather, the rumors in the tavern. The world tells its own story and in games the player walks through it. Eight settings each a different contract with the player. Fantasy, myth and high stakes. Sci-fi, futures that interrogate the present. Post-apocalyptic, survival against ruin. Historical, the past playable. Four more, urban, the familiar estranged. Horror, the world is your guilt. Mythical, dream logic and symbol. Hybrid, every genre at once shifting
[3:15]Under you. Hand the model a setting description and it generates lore entries, environmental narratives, procedural details, all aligned with your world's tone. The setting becomes a prompt that never stops working. Conflict is the engine. No conflict, no challenge, no growth and no reason for the player to act. Every quest, every boss, every hard choice is conflict wearing a costume. Eight conflicts, character versus self, the inner war. Versus character, the rival. Versus society, the unjust system. Versus nature, the cold that doesn't care. Four more, versus technology, the machine's unintended consequences. Versus the supernatural forces beyond understanding versus time, the ticking clock versus the unknown, reality itself
[3:58]In question. For AI, conflict patterns are the quest generator. Feed the model versus society and it writes dilemmas with teeth, choices where both answers cost something. Resolution is the payoff, but not always the answer. It can be closure, ambiguity, or a door that only opens on the second play through. The ending is the last thing the player feels. Seven resolutions closed every thread tied, open-ended the question lingers, branching your choices wrote this ending, circular the loop begins again. Three more, tragic redemption at a price, player defined the ending is wherever you stop, false the closure was a lie, play again. Knowing resolution types lets developers
[4:39]Shape AI-generated endings that stay thematically consistent. The model can improvise the last chapter as long as you've told it what kind of ending this is. Storytelling has always followed its tools. Firelight gave us myths and legends, writing gave us epics, the Iliad, the Ramayana, theater gave us dialogue and an audience to move. Then film and television added image and edit, and games added the missing ingredient, agency. Text adventures then modern games, every play through a co-creation between player and system. Now the AI chapter, AI Dungeon, endless story paths from a prompt, Project December, conversations with emotionally nuanced characters, and modders wiring language models into Skyrim NPCs who
[5:22]Remember what you did. AI doesn't replace the storyteller, it expands what's possible, stories that respond and adapt at a scale no writing team could author by hand. The tools change, the job move us, challenge us, connect us stays. Plot, character, setting, conflict, resolution, the principles don't change when the tools do. Whether the story is written by humans, machines, or both, it still has to move us. The split in six lines. One, AI can't invent story logic. It builds on the narrative structures you give it. Two, plot is the skeleton, classical arcs and eight archetypes. Three, characters are the emotional anchor, seven roles each with a job.
[6:01]Four, setting is a character, too. Eight worlds that shape tone and play. Five, conflict is the engine, eight clashes that generate quests. Six, resolution is the payoff and storytelling's long evolution now has AI as co-author. Your turn. You're designing a game about a lighthouse keeper who finds a door that shouldn't exist. Using the eight plot archetypes, pick the two that fit best and say why. Then sketch a three-act structure and name one quest an AI could generate that still fits your theme. What is storytelling at Nick Bear Brown?
More from Generative AI for Game Development — Shesh Narayan Gupta
4:3305 · Understanding Narrative Structures | Generative AI for Game Development
3:1503 · Purpose and Audience of the Book | Generative AI for Game Development
6:0506 · Types of Interactive Narratives | Generative AI for Game Development
6:3502 · Why Generative AI Matters in Game Development | Generative AI for Game Development
3:3107 · AI Tools and Technologies for Storytelling | Generative AI for Game Development
7:01